Multiple object tracking/recognition

Hi, I’m working on a project to monitor feed on the water surface, using machine learning. So far, the algorithm works on the principle of dividing the image 214x214 into smaller segments 27x27, which it classifies using the model from the edge pulse. It works quite nice but still method has several shortcomings and not as clean as I would like. When looking for a solutions, I found that the edge impulse also allows searching and tracking objects Build Your Own Object Detection System with Machine Learning - YouTube. However, I did not find anything about the implementation with openmv. Is there any way to do this?
I have classic H7 version, not the + one. Model has only 2 classes feed and empty, example of how it works right now:
Thank you :slight_smile:

We will be adding that in the next release. However, we have to finish up the current firmware release. We have a lot of new features about to hit.

1 Like

Thank you very much, I’m looking forward to it, can you reveal in what time range we can expect it? one, two months or more?

Working on getting the current release out this month and a whole bunch of IDE and documentation updates. So, next 2 months afterwards. Expect end of July.

1 Like

Hi, I haven’t watched updates and new releases in a while, so I’m asking here. Did you manage to implement the things mentioned in this thread? Ty :slight_smile:

Yeah, see the Edge Impulse FOMO model.

1 Like

Nice, but i assume, there is no way classic OpenMV H7 would have enough computing power for that. Or ? And don’t you plan to make a tutorial video about it ? It would be great, cuz looks like move to next level for me.

The classic H7 can run it at 30 FPS.

There’s no need for a video. Just go onto Edge Impulses website and try it out.